ASRock Z390 Steel Legend review
Corsair M55 RGB PRO gaming mouse review
Promo: Windows 10 Pro for $13 With Office 2016 For $33
Lexar NM600 NVMe M.2. SSD (480GB) review
Tech preview: Radeon RX 5700 and 5700 XT
Tech preview: AMD Ryzen 3000 - 7nm ZEN2
Patriot Viper VPN100 M2 NVMe 512 GB SSD Review
Samsung 970 EVO Plus 2TB NVMe M.2. SSD review
Guru3D Rig of the Month - May 2019
Promo: URCDKey Summer Sale Windows 10 Pro for $11
Battlefield V: DLSS PC Performance Update
Posted by: on: 02/14/2019 12:06 PM [
57 comment(s) ]
EA and DICE have added DLSS support for Battlefield V. Combined with a new 418.91 driver from NVIDIA we check out the benefits and disadvantages of what is called AI AA in this small first update with Battlefield V.
Read article
Advertisement
pages 1 2 3
Netherwind
Senior Member
Posts: 6519
Senior Member
Posts: 6519
Posted on: 02/14/2019 12:52 PM
What about 3440x1440? Is it supported?
What about 3440x1440? Is it supported?
HWgeek
Senior Member
Posts: 404
Senior Member
Posts: 404
Posted on: 02/14/2019 12:53 PM
thanks Hilbert !
About DLSS, it's not just blurry, it's just has much less details, so of-course the FPS gonna improve, looks like on simple surfaces DLSS can do OK, but on very complicated objects/textured DLSS can't be useful, the affect is just too bad, SRY NV fans :-(.
thanks Hilbert !
About DLSS, it's not just blurry, it's just has much less details, so of-course the FPS gonna improve, looks like on simple surfaces DLSS can do OK, but on very complicated objects/textured DLSS can't be useful, the affect is just too bad, SRY NV fans :-(.
entr0cks
Member
Posts: 64
Member
Posts: 64
Posted on: 02/14/2019 01:19 PM
For what it is I'm surprisingly impressed.
Better article then the previous one.
For what it is I'm surprisingly impressed.
Better article then the previous one.
rsouzadk
Junior Member
Posts: 17
Junior Member
Posts: 17
Posted on: 02/14/2019 01:28 PM
I have an aw3418dw and msi 2080 gaming x trio.
It works on 3440x1440, but the image qualitiy loss doesnt worth it. I do prefer play with dlss+rtx off.
What about 3440x1440? Is it supported?
I have an aw3418dw and msi 2080 gaming x trio.
It works on 3440x1440, but the image qualitiy loss doesnt worth it. I do prefer play with dlss+rtx off.
Fox2232
Senior Member
Posts: 9022
Senior Member
Posts: 9022
Posted on: 02/14/2019 01:36 PM
@Hilbert Hagedoorn : Can we get screenshot with DLSS-ON and "resolution scale" high enough to decrease fps to match DLSS-OFF screenshot?
(So, we can compare IQ at equal fps.)
= = = =
Side note, not allowing DLSS on 4K for weaker cards is like saying: "Pay Extra!"
@Hilbert Hagedoorn : Can we get screenshot with DLSS-ON and "resolution scale" high enough to decrease fps to match DLSS-OFF screenshot?
(So, we can compare IQ at equal fps.)
= = = =
Side note, not allowing DLSS on 4K for weaker cards is like saying: "Pay Extra!"
Undying
Senior Member
Posts: 11035
Senior Member
Posts: 11035
Posted on: 02/14/2019 01:42 PM
Limiting DLSS on different cards per resolution is intentional and disgusting. I agree with Fox, it is like saying pay more.
After such a heavily marketed feature at the end is another useless one.
Limiting DLSS on different cards per resolution is intentional and disgusting. I agree with Fox, it is like saying pay more.
After such a heavily marketed feature at the end is another useless one.
HWgeek
Senior Member
Posts: 404
Senior Member
Posts: 404
Posted on: 02/14/2019 02:05 PM
Little Joke time
.
How would Thanos looked like if he was Renderd with DLSS ON ?
I think I expressed accurately the DLSS effect on quality- am I right?
Little Joke time
.How would Thanos looked like if he was Renderd with DLSS ON ?
I think I expressed accurately the DLSS effect on quality- am I right?
Fox2232
Senior Member
Posts: 9022
Senior Member
Posts: 9022
Posted on: 02/14/2019 02:08 PM
Little Joke time.
How would Thanos looked like if he was Renderd with DLSS ON ?
DLSS turns him from intimidating and angry into elderly man with expression showing trouble to understand something in front of him.
Little Joke time
.How would Thanos looked like if he was Renderd with DLSS ON ?
DLSS turns him from intimidating and angry into elderly man with expression showing trouble to understand something in front of him.
Bigbeard1986
Senior Member
Posts: 102
Senior Member
Posts: 102
Posted on: 02/14/2019 02:28 PM
It is absolutely illogical why i can't use dlss on 1440p on my 2080ti. Geforce experience says it should be on, but it becomes greyed out in game.
Another thing no one ever mentioned...with ray tracing on, i get severe mouse lag.
Anyone else?
It is absolutely illogical why i can't use dlss on 1440p on my 2080ti. Geforce experience says it should be on, but it becomes greyed out in game.
Another thing no one ever mentioned...with ray tracing on, i get severe mouse lag.
Anyone else?
Romulus_ut3
Senior Member
Posts: 658
Senior Member
Posts: 658
Posted on: 02/14/2019 02:49 PM
Hilbert, I'd very much like to see the following:
A comparison between DLSS on/DLSS off/75% resolution scaling at the same resolution/75% resolution scaling at the same resolution on a Radeon VII.
I think that would make for a very interesting comparison.
Hilbert, I'd very much like to see the following:
A comparison between DLSS on/DLSS off/75% resolution scaling at the same resolution/75% resolution scaling at the same resolution on a Radeon VII.
I think that would make for a very interesting comparison.
Rich_Guy
Senior Member
Posts: 12177
Senior Member
Posts: 12177
Posted on: 02/14/2019 03:50 PM
Blury mess, with missing detail, looks terrible.
Blury mess, with missing detail, looks terrible.
Luc
Member
Posts: 48
Member
Posts: 48
Posted on: 02/14/2019 03:58 PM
The upscaling from 1440p to 2160p hits the performance in 10 fps, from 65 to 55... so now I'm interested in seeing the differences between 1440p and DLSS-2160p, triying to find where to blame Nvidia :p
Btw, great work Hilbert, I like the sliders a lot, another toy to lose time with.
The upscaling from 1440p to 2160p hits the performance in 10 fps, from 65 to 55... so now I'm interested in seeing the differences between 1440p and DLSS-2160p, triying to find where to blame Nvidia :p
Btw, great work Hilbert, I like the sliders a lot, another toy to lose time with.
Valken
Senior Member
Posts: 1451
Senior Member
Posts: 1451
Posted on: 02/14/2019 04:26 PM
Nvidia RTX - features set that keeps on pissing off RTX owners...
Nvidia RTX - features set that keeps on pissing off RTX owners...
waltc3
Senior Member
Posts: 926
Senior Member
Posts: 926
Posted on: 02/14/2019 05:05 PM
Enabling DLSS only for 4k resolutions makes no sense, obviously...but perhaps it looks a lot less blurry @ 4k--that would be my guess here. Restricting its use is certainly is no oversight. Ah, yes, the nVidia I remember so well is back! Pulling image-quality tricks, deceits, and sleights of hand, once again...! It's Back to the Future, alright....
It's indeed gratifying to know that when blurriness increases so does frame-rate--even if the Tensor cores kill the frame-rate by nearly half. (That in itself is weird from nVidia, because the company has traditionally been all about benchmarked frame rates and very little else--fought for years against FSAA until they learned how to do it in hardware, etc.)
The one thing I didn't really understand here is the difference in the comparative screen shots--the DLSS On shots show vegetation close up, while the DLSS Off screens put the vegetation in the far distance and obscured by fog. Scratching my head on that one...?
Enabling DLSS only for 4k resolutions makes no sense, obviously...but perhaps it looks a lot less blurry @ 4k--that would be my guess here. Restricting its use is certainly is no oversight. Ah, yes, the nVidia I remember so well is back! Pulling image-quality tricks, deceits, and sleights of hand, once again...! It's Back to the Future, alright....
It's indeed gratifying to know that when blurriness increases so does frame-rate--even if the Tensor cores kill the frame-rate by nearly half. (That in itself is weird from nVidia, because the company has traditionally been all about benchmarked frame rates and very little else--fought for years against FSAA until they learned how to do it in hardware, etc.) The one thing I didn't really understand here is the difference in the comparative screen shots--the DLSS On shots show vegetation close up, while the DLSS Off screens put the vegetation in the far distance and obscured by fog. Scratching my head on that one...?
Denial
Senior Member
Posts: 12117
Senior Member
Posts: 12117
Posted on: 02/14/2019 05:12 PM
Enabling DLSS only for 4k resolutions makes no sense, obviously...but perhaps it looks a lot less blurry @ 4k--that would be my guess here. Restricting its use is certainly is no oversight. Ah, yes, the nVidia I remember so well is back! Pulling image-quality tricks, deceits, and sleights of hand, once again...! It's Back to the Future, alright.... It's indeed gratifying to know that when blurriness increases so does frame-rate--even if the Tensor cores kill the frame-rate by nearly half. (That in itself is weird from nVidia, because the company has traditionally been all about benchmarked frame rates and very little else--fought for years against FSAA until they learned how to do it in hardware, etc.)
The one thing I didn't really understand here is the difference in the comparative screen shots--the DLSS On shots show vegetation close up, while the DLSS Off screens put the vegetation in the far distance and obscured by fog. Scratching my head on that one...?
DLSS isn't restricted to 4K in BF5?
Also you keep confusing Tensor cores with RT cores. Tensor does DLSS, DLSS does not "cut frame-rate by half" RT cores do RTX/DXR - that does cut framerate. Nvidia intended Tensors to denoise the RT image but DICE doesn't use tensors at all for RT in BF5.
Enabling DLSS only for 4k resolutions makes no sense, obviously...but perhaps it looks a lot less blurry @ 4k--that would be my guess here. Restricting its use is certainly is no oversight. Ah, yes, the nVidia I remember so well is back! Pulling image-quality tricks, deceits, and sleights of hand, once again...! It's Back to the Future, alright....
It's indeed gratifying to know that when blurriness increases so does frame-rate--even if the Tensor cores kill the frame-rate by nearly half. (That in itself is weird from nVidia, because the company has traditionally been all about benchmarked frame rates and very little else--fought for years against FSAA until they learned how to do it in hardware, etc.)The one thing I didn't really understand here is the difference in the comparative screen shots--the DLSS On shots show vegetation close up, while the DLSS Off screens put the vegetation in the far distance and obscured by fog. Scratching my head on that one...?
DLSS isn't restricted to 4K in BF5?
Also you keep confusing Tensor cores with RT cores. Tensor does DLSS, DLSS does not "cut frame-rate by half" RT cores do RTX/DXR - that does cut framerate. Nvidia intended Tensors to denoise the RT image but DICE doesn't use tensors at all for RT in BF5.
HWgeek
Senior Member
Posts: 404
Senior Member
Posts: 404
Posted on: 02/14/2019 05:38 PM
My main complaint on DLSS is that if nVIdia chose to market it like "Better Quality at same Performance" and indeed Enabling DLSS in game resulted as Deep Learning Up-scaled image with much better quality without decreasing performance- That's Great and worth the $$, But using fancy Tech that Reduces Image Quality to Improve Performance? If I am note mistaken- this was called "Cheating" back in the days.
Also Limiting the options of when and how you can use DLSS just proves my point.
My main complaint on DLSS is that if nVIdia chose to market it like "Better Quality at same Performance" and indeed Enabling DLSS in game resulted as Deep Learning Up-scaled image with much better quality without decreasing performance- That's Great and worth the $$, But using fancy Tech that Reduces Image Quality to Improve Performance? If I am note mistaken- this was called "Cheating" back in the days.
Also Limiting the options of when and how you can use DLSS just proves my point.
Denial
Senior Member
Posts: 12117
Senior Member
Posts: 12117
Posted on: 02/14/2019 05:59 PM
My main complaint on DLSS is that if nVIdia chose to market it like "Better Quality at same Performance" and indeed Enabling DLSS in game resulted as Deep Learning Up-scaled image with much better quality without decreasing performance- That's Great and worth the $$, But using fancy Tech that Reduces Image Quality to Improve Performance? If I am note mistaken- this was called "Cheating" back in the days.
Also Limiting the options of when and how you can use DLSS just proves my point.
Well they specifically marketed as an improved TAA. If you go and read the Siggraph whitepaper/slide from 2017 on the early versions of DLSS (Convolutional Auto-Encoder) it was basically designed as a TAA variant that gets rid of the some of the TAA motion artifact issues.
The thing about TAA though is that it's not a standard solution. Each dev integrates it differently with varying degrees of quality. FFXV for example has terrible TAA, same with Port Royal. DICE typically does a good job with their variant but they are also notorious for not allowing you to disable it. Further the issues with TAA are mostly motion issues so people comparing single frames of games aren't really looking for the things DLSS was supposed to solve in the first place.
The limitations of how/when you can use DLSS is more of an intrinsic property of how DLSS training works then nefariousness by Nvidia. They spend weeks training each resolution and RTX On vs Off has to be done in separate training sessions. Should also point out that the training is ongoing and Nvidia says they'll be delivering DLSS updates to games via Experience.
Also I don't think an option to turn something on that can potentially reduce quality for performance improvement is considered cheating.
My main complaint on DLSS is that if nVIdia chose to market it like "Better Quality at same Performance" and indeed Enabling DLSS in game resulted as Deep Learning Up-scaled image with much better quality without decreasing performance- That's Great and worth the $$, But using fancy Tech that Reduces Image Quality to Improve Performance? If I am note mistaken- this was called "Cheating" back in the days.
Also Limiting the options of when and how you can use DLSS just proves my point.
Well they specifically marketed as an improved TAA. If you go and read the Siggraph whitepaper/slide from 2017 on the early versions of DLSS (Convolutional Auto-Encoder) it was basically designed as a TAA variant that gets rid of the some of the TAA motion artifact issues.
The thing about TAA though is that it's not a standard solution. Each dev integrates it differently with varying degrees of quality. FFXV for example has terrible TAA, same with Port Royal. DICE typically does a good job with their variant but they are also notorious for not allowing you to disable it. Further the issues with TAA are mostly motion issues so people comparing single frames of games aren't really looking for the things DLSS was supposed to solve in the first place.
The limitations of how/when you can use DLSS is more of an intrinsic property of how DLSS training works then nefariousness by Nvidia. They spend weeks training each resolution and RTX On vs Off has to be done in separate training sessions. Should also point out that the training is ongoing and Nvidia says they'll be delivering DLSS updates to games via Experience.
Also I don't think an option to turn something on that can potentially reduce quality for performance improvement is considered cheating.
HWgeek
Senior Member
Posts: 404
Senior Member
Posts: 404
Posted on: 02/14/2019 06:08 PM
Thank for the info , I do hope that in coming months we gonna see better results.
I am just worried that NV will go AI/DL route and claim FPS boost with the technology instead of improving the silicone (for example boosting performance between GTX 680/780/980/1080 ).
Thank for the info , I do hope that in coming months we gonna see better results.
I am just worried that NV will go AI/DL route and claim FPS boost with the technology instead of improving the silicone (for example boosting performance between GTX 680/780/980/1080 ).
leszy
Senior Member
Posts: 313
Senior Member
Posts: 313
Posted on: 02/14/2019 06:09 PM
We all know that bitmaps can not be enlarged arbitrarily, because even the most perfect filter can not extract/display sub-pixel information, because it has not been saved (it simply does not exist). The same applies to upscaling. You can create the illusion of sharpness at first glance, but no anti-aliasing technique will allow you to display details that have not been rendered due to the low resolution of the rendered frame. Perhaps it's worth checking whether, in general, using high resolution textures makes sense with DLSS enabled? Maybe smaller textures will not worsen the quality of the image and will boost fps?
We all know that bitmaps can not be enlarged arbitrarily, because even the most perfect filter can not extract/display sub-pixel information, because it has not been saved (it simply does not exist). The same applies to upscaling. You can create the illusion of sharpness at first glance, but no anti-aliasing technique will allow you to display details that have not been rendered due to the low resolution of the rendered frame. Perhaps it's worth checking whether, in general, using high resolution textures makes sense with DLSS enabled? Maybe smaller textures will not worsen the quality of the image and will boost fps?
HardwareCaps
Senior Member
Posts: 452
Senior Member
Posts: 452
Posted on: 02/14/2019 06:18 PM
Well they specifically marketed as an improved TAA. If you go and read the Siggraph whitepaper/slide from 2017 on the early versions of DLSS (Convolutional Auto-Encoder) it was basically designed as a TAA variant that gets rid of the some of the TAA motion artifact issues.
The thing about TAA though is that it's not a standard solution. Each dev integrates it differently with varying degrees of quality. FFXV for example has terrible TAA, same with Port Royal. DICE typically does a good job with their variant but they are also notorious for not allowing you to disable it. Further the issues with TAA are mostly motion issues so people comparing single frames of games aren't really looking for the things DLSS was supposed to solve in the first place.
The limitations of how/when you can use DLSS is more of an intrinsic property of how DLSS training works then nefariousness by Nvidia. They spend weeks training each resolution and RTX On vs Off has to be done in separate training sessions. Should also point out that the training is ongoing and Nvidia says they'll be delivering DLSS updates to games via Experience.
Also I don't think an option to turn something on that can potentially reduce quality for performance improvement is considered cheating.
New anti-aliasing techniques are always welcomed, giving the user freedom of choice and flexibility.
The issue is that DLSS requires dedicated hardware and is only supported by a single architecture(so far). you're basically paying in terms of thermals, die-size, transistors for DLSS.
if DLSS doesn't prove itself to be very close in terms of visuals then it makes no sense, that die-space can be used instead for CUDA cores that will negate the performance "increase" gained by DLSS
while still being 100% useful on all available games.
Well they specifically marketed as an improved TAA. If you go and read the Siggraph whitepaper/slide from 2017 on the early versions of DLSS (Convolutional Auto-Encoder) it was basically designed as a TAA variant that gets rid of the some of the TAA motion artifact issues.
The thing about TAA though is that it's not a standard solution. Each dev integrates it differently with varying degrees of quality. FFXV for example has terrible TAA, same with Port Royal. DICE typically does a good job with their variant but they are also notorious for not allowing you to disable it. Further the issues with TAA are mostly motion issues so people comparing single frames of games aren't really looking for the things DLSS was supposed to solve in the first place.
The limitations of how/when you can use DLSS is more of an intrinsic property of how DLSS training works then nefariousness by Nvidia. They spend weeks training each resolution and RTX On vs Off has to be done in separate training sessions. Should also point out that the training is ongoing and Nvidia says they'll be delivering DLSS updates to games via Experience.
Also I don't think an option to turn something on that can potentially reduce quality for performance improvement is considered cheating.
New anti-aliasing techniques are always welcomed, giving the user freedom of choice and flexibility.
The issue is that DLSS requires dedicated hardware and is only supported by a single architecture(so far). you're basically paying in terms of thermals, die-size, transistors for DLSS.
if DLSS doesn't prove itself to be very close in terms of visuals then it makes no sense, that die-space can be used instead for CUDA cores that will negate the performance "increase" gained by DLSS
while still being 100% useful on all available games.
Denial
Senior Member
Posts: 12117
Senior Member
Posts: 12117
Posted on: 02/14/2019 06:57 PM
New anti-aliasing techniques are always welcomed, giving the user freedom of choice and flexibility.
The issue is that DLSS requires dedicated hardware and is only supported by a single architecture(so far). you're basically paying in terms of thermals, die-size, transistors for DLSS.
if DLSS doesn't prove itself to be very close in terms of visuals then it makes no sense, that diespace can be used for CUDA cores who will negate the perfomance "increase" gained by DLSS
while still being 100% useful on all available games.
Volta is essentially the same architecture as Turing and Titan V has significantly more CUDA cores than a 2080Ti yet performs identically (even when both are overclocked) because the entire architecture is essentially TDP limited. Regardless, Tensor 'cores' exist within the SM - they aren't discreet cores but operational modes of SM/ALUs - same as RT 'cores'. They add some negligible amount of die space but for the most part they reuse existing transistors within the SM. People keep alluding to the fact that they take up all this die space but no one has been able to explain to me how Nvidia managed to get the same CUDA/MM2 density as GP100, on the same process density, with double the cache, dispatch units and RT/Tensor that "take up a ton of space".
Further AMD exposed INT4/8 for ML inferencing on Vega 7 presumably to support some type of GPUOpen variant of DLSS on DirectML - which they alluded to doing. DirectML support's tensors so even if there was a vendor agnostic standard it would still utilize the hardware.
New anti-aliasing techniques are always welcomed, giving the user freedom of choice and flexibility.
The issue is that DLSS requires dedicated hardware and is only supported by a single architecture(so far). you're basically paying in terms of thermals, die-size, transistors for DLSS.
if DLSS doesn't prove itself to be very close in terms of visuals then it makes no sense, that diespace can be used for CUDA cores who will negate the perfomance "increase" gained by DLSS
while still being 100% useful on all available games.
Volta is essentially the same architecture as Turing and Titan V has significantly more CUDA cores than a 2080Ti yet performs identically (even when both are overclocked) because the entire architecture is essentially TDP limited. Regardless, Tensor 'cores' exist within the SM - they aren't discreet cores but operational modes of SM/ALUs - same as RT 'cores'. They add some negligible amount of die space but for the most part they reuse existing transistors within the SM. People keep alluding to the fact that they take up all this die space but no one has been able to explain to me how Nvidia managed to get the same CUDA/MM2 density as GP100, on the same process density, with double the cache, dispatch units and RT/Tensor that "take up a ton of space".
Further AMD exposed INT4/8 for ML inferencing on Vega 7 presumably to support some type of GPUOpen variant of DLSS on DirectML - which they alluded to doing. DirectML support's tensors so even if there was a vendor agnostic standard it would still utilize the hardware.
MonstroMart
Senior Member
Posts: 420
Senior Member
Posts: 420
Posted on: 02/14/2019 07:07 PM
Going by the screenshots posted i think it looks better with DLSS off. Looks less detailed and slightly blurry with DLSS. Looks like the game is running in an inferior resolution than the native monitor resolution and upscaled with a good upscaler. No reason to do that if you are running an expensive 4k monitor. Might as well spend the extra money to get a 2080TI and run the game without DLSS.
Going by the screenshots posted i think it looks better with DLSS off. Looks less detailed and slightly blurry with DLSS. Looks like the game is running in an inferior resolution than the native monitor resolution and upscaled with a good upscaler. No reason to do that if you are running an expensive 4k monitor. Might as well spend the extra money to get a 2080TI and run the game without DLSS.
HardwareCaps
Senior Member
Posts: 452
Senior Member
Posts: 452
Posted on: 02/14/2019 07:25 PM
Volta is essentially the same architecture as Turing and Titan V has significantly more CUDA cores than a 2080Ti yet performs identically (even when both are overclocked) because the entire architecture is essentially TDP limited. Regardless, Tensor 'cores' exist within the SM - they aren't discreet cores but operational modes of SM/ALUs - same as RT 'cores'. They add some negligible amount of die space but for the most part they reuse existing transistors within the SM. People keep alluding to the fact that they take up all this die space but no one has been able to explain to me how Nvidia managed to get the same CUDA/MM2 density as GP100, on the same process density, with double the cache, dispatch units and RT/Tensor that "take up a ton of space".
Further AMD exposed INT4/8 for ML inferencing on Vega 7 presumably to support some type of GPUOpen variant of DLSS on DirectML - which they alluded to doing. DirectML support's tensors so even if there was a vendor agnostic standard it would still utilize the hardware.
not sure where you got that info(nvidia didn't reveal much) but in their blog they clearly state that Tensor cores populate a decent amount out of the SMs while RT cores take additional die-space.
they managed to get more cache in because of the fact that the die-sizes are much bigger on Turing(also 12nm does improve density slightly)
https://devblogs.nvidia.com/nvidia-turing-architecture-in-depth/
Also a clear indication is that transistor count has grown between Pascal and Turing on the same CUDA count.
GTX 2060 - https://www.techpowerup.com/gpu-specs/geforce-rtx-2060.c3310
10,800 transistors, 445 mm die
GTX 1070 - https://www.techpowerup.com/gpu-specs/geforce-gtx-1070.c2840
7,200 transistors, 314 mm die
huge difference
Volta is essentially the same architecture as Turing and Titan V has significantly more CUDA cores than a 2080Ti yet performs identically (even when both are overclocked) because the entire architecture is essentially TDP limited. Regardless, Tensor 'cores' exist within the SM - they aren't discreet cores but operational modes of SM/ALUs - same as RT 'cores'. They add some negligible amount of die space but for the most part they reuse existing transistors within the SM. People keep alluding to the fact that they take up all this die space but no one has been able to explain to me how Nvidia managed to get the same CUDA/MM2 density as GP100, on the same process density, with double the cache, dispatch units and RT/Tensor that "take up a ton of space".
Further AMD exposed INT4/8 for ML inferencing on Vega 7 presumably to support some type of GPUOpen variant of DLSS on DirectML - which they alluded to doing. DirectML support's tensors so even if there was a vendor agnostic standard it would still utilize the hardware.
not sure where you got that info(nvidia didn't reveal much) but in their blog they clearly state that Tensor cores populate a decent amount out of the SMs while RT cores take additional die-space.
they managed to get more cache in because of the fact that the die-sizes are much bigger on Turing(also 12nm does improve density slightly)
https://devblogs.nvidia.com/nvidia-turing-architecture-in-depth/
Also a clear indication is that transistor count has grown between Pascal and Turing on the same CUDA count.
GTX 2060 - https://www.techpowerup.com/gpu-specs/geforce-rtx-2060.c3310
10,800 transistors, 445 mm die
GTX 1070 - https://www.techpowerup.com/gpu-specs/geforce-gtx-1070.c2840
7,200 transistors, 314 mm die
huge difference
Mesab67
Member
Posts: 65
Member
Posts: 65
Posted on: 02/14/2019 07:49 PM
Thanks for the update Hilbert.
I don't buy this. Any feature that effectively lowers final output resolution, particularly of textures, just to lower processing cost/meet fps acceptability is a worrying trend. What's the point in developing high/"ultra" res textures (/later 'texture packs') in the first place? I'm sure the game artists will getting pretty pissed off.
We see the same with the RT feature - scene RT complexity reduction to meet fps targets. I do wish Nvidia placed a higher weighting on their QA department than marketing...get your act together Nvidia!!
I purchased a 1440p monitor for one primary reason - to be able to benefit from higher resolution, crisp graphics. I don't want the artists hard work smeared.
Thanks for the update Hilbert.
I don't buy this. Any feature that effectively lowers final output resolution, particularly of textures, just to lower processing cost/meet fps acceptability is a worrying trend. What's the point in developing high/"ultra" res textures (/later 'texture packs') in the first place? I'm sure the game artists will getting pretty pissed off.
We see the same with the RT feature - scene RT complexity reduction to meet fps targets. I do wish Nvidia placed a higher weighting on their QA department than marketing...get your act together Nvidia!!
I purchased a 1440p monitor for one primary reason - to be able to benefit from higher resolution, crisp graphics. I don't want the artists hard work smeared.
pages 1 2 3
Click here to post a comment for this article on the message forum.
Senior Member
Posts: 452
Great job adding the comparisons!
DLSS is definitely more blurry, look at the vegetation and the trees. so blurry for a 4K output....
also the vehicle on fire is far more detailed with native 4K